A Stochastic Optimization Model for Real‐Time Operation of Reservoirs Using Uncertain Forecasts

Publisher Website
Google Scholar
Abstract
A real‐time operation model primarily useful for daily operation of reservoirs is developed. This model is based on a chance constraint formulation and assumes a particular form of the linear decision rule. It uses the conditional distribution function (CDF) of actual streamflows conditioned on the forecasted values. These CDF's are constructed by incorporating the statistical properties of forecast errors for different time steps. The objective considered is the minimization of weighted probable deviations from storage and release targets. These weights are surrogates for the actual loss functions, and the probable deviations are functions of the reliability levels specified in the model. With the use of target values for release and storage, this model is capable of using a release policy that is a subset of a seasonal policy and overcomes the myopic (short‐sighted) nature of operation. Simulation of actual operation, using this model for a hypothetical reservoir, demonstrates the feasibility and efficiency of this approach. This model is shown to be applicable for a system of reservoirs. The restrictions associated with the use of a linear decision rule are shown to be invalid for this model.